| Abstrakt: |
Dividend policy is one of the most discussed and controversial topics in corporate finance for decades. Due to the increase of computational power, few scholars tried to find most informative determinants with the help of advanced heuristic methods, a.k.a. Machine Learning. However, some critiques need to be addressed regarding the metric scores, model selection or robustness of their approaches. This paper proposes a working methodology that deals with these critiques. A Principal Components Analysis as well as numerous ML models, resample techniques and metric scores have been applied in order to better understand the dividend puzzle. The conclusions suggest that the size of the companies is the most informative determinant, larger and less risky firms being more likely to pay dividends. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink